Research On Data-driven Based Identification Method Of Hollow Slab Bridge’s Joint Damage Under Driving Vehicle’s Excitation

As a common bridge type in small and medium-sized bridges,prefabricated hollow slab bridges are often caused by various external effects during the design,construction and operation.The hinge joints often appear local damage or failure before the hollow slab girder,which will greatly reduce the horizontal integrity of the bridge.Causing great hidden dangers and affecting the safety of the bridge.Therefore,it is necessary to monitor the health status of joints of hollow slab bridges in service in real time.However,due to the highly non-linear implicit function relationship between the dynamic response of the bridge under normal traffic loads and the joint state,it is impossible to directly judge the health state of each joint directly through the dynamic response of the bridge.Therefore,this paper carries out research on Datadriven based identification method of hollow slab bridge’s joint damage under driving vehicle’s excitation.Firstly,selecting the hinge joints damage numerical simulation method based on the mechanical characteristics of the hollow slab bridge and the damage characteristics of hinge joints.Secondly,based on the mutual calling of commercial software ANSYS and MATLAB and the overall iterative method,the entity refined vehicle-bridge coupled vibration analysis system was established,select representative highway first-class vehicle loads and typical three-axle vehicles as the loaded models and explore the most vulnerable typical hinge joints positions under normal operation,and make a preliminary classification of the joint damage levels.Finally,the vehicle type,vehicle type,and driving lane are used as control variables to analyze the dynamic response of the bridge with different joint damage under a variety of influencing parameter conditions,and the Pearson similarity damage index is proposed and used as a data driven,using the gradient to improve the decision tree algorithm training Damage identification model,realizing the location and quantitative description of hinge damage.The conclusions are as follows:(1)Based on the existing typical hollow slab bridge hinge joints structure form,analyze the causes of hinge joints failure from the design,construction and operation aspects,and based on the transverse force characteristics of the hollow slab bridge and the hinge joint dilapidation process,compare the modified transverse stiffness method of the beam lattice model and solid model deduction element method on the simulation consequent of hinge joint damage.By comparison,it is indicated that the solid model deduction element method is closer to the actual force of the hollow slab bridge.The influence of the bridge deck integration layer on the lateral force of the bridge after the hinge joint damage can be considered,and by deducting different quantity of hinge joint elements,the process of hinge joint damage extending from the bottom to the top can be simulated.Therefore,choosing the solid model to simulate the hinge joint damage is more in line with reality and the calculation accuracy is higher.(2)Based on the refined physical vehicle-bridge coupling analysis,it is found that under the normal multi-lane vehicle load,the stress state of hinge joints at different lateral positions is quite different,and there are significant differences in the failure probability of hinge joints.Generally speaking,hinge joints near the carriageway are more vulnerable.(3)Based on the dynamic response of each slab span under different degrees of hinge joint damage,it can be found that the degree of hinge joints damage increases,the deflection amplitude and acceleration amplitude of adjacent plates increase significantly,and the acceleration time history of adjacent plates increases.The similarity gradually weakened.Based on this,a Pearson similarity damage characteristic index is proposed to characterize the correlation of the dynamic response of adjacent plates.As the degree of hinge joints damage continues to increase,the Pearson similarity damage characteristic value gradually decreases.However,since there is still a highly non-linear mapping relationship between the degree of hinge joints damage and the similarity damage characteristic value,this index is only suitable for hinge joints damage location,and quantitative work cannot be carried out.(4)With simulation and analysis of the dynamic response of hollow slab bridges under different vehicle types,different vehicle speeds and different driving lanes,according to comparative analysis of the impact of various parameters of the hinge joints on the Pearson similarity damage index under the same degree of damage of the vehicle type,vehicle speed and lane position,our results show that under the same damage degree of the hinge joints,the Pearson similarity damage characteristic index changes due to the influence of the model,the speed of the vehicle and the driving lane.Therefore,in order to achieve a quantitative description of hinge damage,the influence of these three factors must be considered comprehensively.(5)Taking the dynamic response of different vehicle types,vehicle speeds,and driving lanes as the original data,a large number of Pearson similarity damage feature indicators are obtained through MATLAB data processing.Input these data as sample of training,a method of joints damage identification based on GBDT algorithm is proposed.With numerical simulation verification,the prediction model obtained by training has high prediction accuracy,and can basically realize the positioning and quantitative description of hinge joints damage;The anti-noise capability of this method is investigated under different noise levels,and the results indicate that the prediction model has strong anti-noise ability and good robustness.(6)Based on the same sample library,this method is compared and analyzed by training traditional machine learning algorithms.The results show that the damage recognition effect of the GBDT algorithm is significantly better than that of the traditional machine learning method.Its generalization ability is stronger,and the fitting accuracy is improved to a certain extent.To sum up,the hinge joints damage identification method based on GBDT training can basically realize the quantitative description of hinge joints damage,and has the potential to be applied in actual engineering,but it still needs to be conducted in-depth research and perfection on the performance under the coupling effect of multiple environmental factors.